cadastral applications, the distance to the remote receiver 
should not be more than 15-20 km. The only potential error 
source which has not been extensively studied are 
differential rotations between the INS and the remote 
sensing device. In other words, to achieve the high attitude 
accuracy, flexing of the aircraft hull, as well as vibration 
differences at the two locations have to be investigated. In 
principle, this amounts to the question whether or not the 
: : b : : t : 
rotation matrix dR, can be considered as time invariant or 
not. If the differential rotations are well below the noise 
level of 15"-30", current high precision INS together with 
DGPS can be used to meet the attitude requirements of even 
the most demanding applications discussed in this paper. 
The determination of the translation vector drP is not as 
critical. Even if the length of the vector changes somewhat 
in time, it will usually stay well within the accuracy level of 
5-10 cm. If photogrammetric cameras are used for high 
precision applications and a block adjustment is possible, 
the georeferencing problem can be reduced to precise 
positioning of the aircraft. This can be done with sufficient 
accuracy by using differential GPS and carrier phase 
techniques. In this case, the costs for the georeferencing 
system can be considerably reduced. 
If high accuracy applications are excluded, more economic 
solutions to the georeferencing problem can be found. Such 
systems would be suitable for the bulk of applications in the 
resource sector. They will therefore most likely be the 
standard georeferencing systems of the future.Two solutions 
seem to be especially attractive. One is completely based on 
GPS technology, the other is a GPS/INS integration with a 
lower cost INS. 
The GPS solution would combine two narrow-correlator C/A 
code receivers for positioning with a GPS multi-antenna 
system for attitude determination. The system would be low 
cost and has the advantage that the same receivers can be 
used for both tasks and sufficient redundancy can be built 
into the system design. Drawbacks are system stability and 
data rate. To get full attitude resolution, one or two of the 
antennas have to be installed on the wings. The resulting 
attitude is affected by wing flutter and most likely deviates 
considerably from the attitude of the remote sensing system. 
This problem is aggravated by the fact that a 50 Hz output 
rate is needed. Current GPS output rates are at 2 Hz, thus 
interpolation is needed. The compounded influence of the 
two effects has to be investigated to confirm the suitability 
of such a system for the stated accuracy requirements. The 
system would not be suitable for SAR-type applications 
because the short-term velocity accuracy is not sufficient. 
The medium cost GPS/INS integration solves the above 
problems and has sufficient short-term velocity resolution 
to also be used in SAR applications. It covers therefore a 
rather broad range of applications. Since the INS can be 
mounted on the same frame or the same platform as the 
imaging system, there is no problem with system stability, 
as long as the rigidity of the frame under vibrations is 
carefully checked. Output rate is not a problem because 
inertial systems come with rates of 50-100 Hz anyway. A 
further advantage is that an onboard INS can usually be 
modified to form part of the integrated system. This will 
reduce the costs considerably. It appears therefore that the 
development of such a system offers the best balance 
between economy and technological risk. This is one of the 
major reasons why the prototype development at the U of C 
moved in this direction. 
7. TESTING OF THE U OF C PROTOTYPE 
SYSTEM 
The current prototype system at the U of C consists of a 
strapdown INS of the medium accuracy class (Litton LTN 
90/100) and two Ashtech P XII receivers. This system was 
recently tested with the Compact Airborne Spectrographic 
Imager (casi) developed by Itres Research Ltd for this 
project (Babey & Anger, 1989). The casi is a pushbroom 
sensor that acquires one scanline at a time as it travels along 
the flight line. The resulting image possesses a different set 
of position and orientation parameters for each scanline, and 
it often contains large distortions induced by movements on 
the aircraft. An example of such an image is shown in Figure 
3: 
The calibration parameters in Equation (5) are f, Xp: Jp Xy, 
b 
and dR, and drb. Al parameters, except drP were 
determined from a calibration target array in the area flown, 
using a self-calibrating bundle adjustment. The orientation 
differences between INS and the imaging sensor, an? will 
change each time the imager and INS is installed in the 
b net 
aircraft. In fact, dR, may change during flight if the camera 
is not rigidly secured with respect to the INS. However, 
b ; : : 
dR, can be calibrated from the imagery if three or more 
control points are present in the block. A block is defined 
by a series of overlapping flightlines. For convenience four 
points were chosen for calibration. A bundle adjustment is 
used to solve for these parameters, Gibson & Buchheit 
(1990). 
To fully geocorrect the imagery, Cosandier et al, (1992) and 
create an ortho-image, the ground height must be taken into 
account via a digital elevation model (DEM). The 
integration and extraction of a DEM from casi is currently 
being developed. 
  
Figure 3: Raw Pushbroom Image 
198 
  
  
7.1. Calg 
À special 
Research Li 
nadir and a 
The sensor | 
view of 43. 
nadir. Thes 
geometry. 
Three flight 
flight line v 
size of 1.6 
metres. Inc 
a Litton L 
receiver. Al 
the test are: 
accuracy. 
Due to the f 
of the area, 
Therefore, : 
instead. T! 
errors in ma 
to the imag 
would be th 
7.2 Test 
To test the : 
are shown. 
accuracies 
adjustment 
The values 
compared fc 
Three differ 
data set. Th 
d) First, t 
points 
gives tl 
used to 
b) Next, t 
points 
corresp 
calibra! 
as che 
scenar 
C) Lastly, 
points 
remain 
This sc 
spectra 
Test (a) res 
about 1.2 | 
error was 5 
are due par 
Standard e 
registration